Raul asked: > Does anyone else have any thoughts on this sort of boundary condition > issue? > There are many problems where the neighbours of cells convey important information for an algorithm. The solution which you listed using 'pad' and 'padded' approaches the problem in a systematic way which separates the padding and the tesselation stages of the process, and for many problems that provides a range of alternatives which can be easily handled within the existing tools.
Separating the two stages simplifies the thinking and provides a greater range of alternatives. These problems arise frequently with non-parametric smoothing, and the existing tools seem to me to be apt. Fraser ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
